May 11, 2016

Baseball Therapy

The Even Slightly More Convincing Argument Against the Shift

by Russell A. Carleton

Last week, we looked into The Shift and whether it was actually doing what we said it was supposed to do, which is to be a better way of getting hitters, especially pull-happy hitters (and double especially groundball-heavy, pull-happy, left-handed hitters) to make more outs. The traditional story of The Shift is that because those hitters are going to be sending most of their ground balls to one side of the field, why not put more fielders over that way?

But an interesting thing happened when we looked at the numbers. I found that players actually had a (slightly) better BABIP, though a (slightly) worse SLG on contact in shifted plate appearances than we would have otherwise expected by looking at their BABIP in non-shifted situations. That means that The Shift was actually producing fewer outs than we might have expected than the non-shift baseline. Now, that’s in the aggregate and league-wide, but it’s surprising for something that’s supposed to be a net positive. Doing a little more research, I found that in shifted situations, hitters produce more grounders than we might expect of them and go oppo (slightly) more often than we would expect. The Shift might be better at gobbling up the grounders that are hit, but it won’t get all of them, and some of those grounders would have otherwise been flyballs, which are much easier to catch. The net effect is at best neutral.

It’s a piece that has provoked some strong reactions. No less a luminary than Ben Jedlovec of Baseball Info Solutions (who provided the data that I used in my original work) proposed that one flaw in my research methodology was that because I set my minimum inclusion criteria at 100 non-shifted PA’s in the year in question, it left out the players who got shifted the most. David Ortiz doesn’t get 100 non-shifted PA’s in a season, so he’s de facto left out of the sample, and he might be the quintessential shifted player.

Mr. Jedlovec essentially duplicated my methodology, but relaxed those inclusion criteria a little bit and found a (small) net positive effect for The Shift when it comes to BABIP. His methodology makes sense, although I think it’s telling that the Shift’s positive effects can only be re-established when the most “shift-able” guys are thrown back into the mix. It seems that what we have are some guys who are very shift-able (hello, Mr. Ortiz!) and when they are taken out, The Shift becomes a net negative. I take that to mean that teams shift against far too many hitters. It’s not that they should take The Shift off for everyone, or even for everyone except the David Ortizes of the world, but overall, there are a lot of shifts that are being done that are actually counter-productive, if only slightly so.

And then there’s this. These are analyses that I conducted last week, but didn’t publish (mostly for the reason that the piece was already 3,000 words long by the time I got there.)

I should probably…

Warning! Gory Mathematical Details Ahead!

I essentially ran my analyses from last week backwards. I looked at all hitters who had been shifted against at least 50 times during the year in question. This will most certainly get the guys who are heavily shifted into the sample. I looked at their shifted stats as my baseline and looked at what happened when they weren’t shifted. Again, this is summed across the entire league.

Year	No-Shift Hits (expected)	No-Shift Hits (actual)	No-Shift Total Bases (expected)	No-Shift Total Bases (actual)
2012	1,714	1,696	2,206	2,239
2013	3,091	3,290	3,966	4,274
2014	4,310	4,157	5,464	5,422
2015	5,264	5,321	6,772	6,888
Total	14,379	14,464	18,408	18,823

In this case, if The Shift really is superior, we should see that when teams switch to an inferior defense against these same hitters, they should increase the number of hits that they get compared to expectations. And they do. In odd-numbered years. The overall effect is (slightly) net positive toward the defense, but not the kind of numbers that would make anyone stand up and cheer. At the very best—in the aggregate and as the shift is practiced now—The Shift is a net neutral when it comes to defensing balls in play.

We’re used to looking at the list of teams who shift the most and patting them on the head for being so forward-thinking. They may very well have been the teams that initially found the magic hammer but, it seems that they have made the mistake of running around and using it to pound all of the existent and non-existent nails that they see.

***

When I opened the piece last week, I began it with what I referred to as a “cheeky question” which was how many home runs had been hit against The Shift. It seems like a silly question because if the batter hits a 500 foot blast, there’s nothing that The Shift is going to do to bring it back into the park. But still, I wondered whether The Shift might make pitchers throw differently, perhaps “taking away” one of their pitches or a location and making the pitchers more predictable. That might make it easier to tee off on him. Maybe strikeouts and walks would also be affected.

The way that the data which are publicly available are set up, there’s no accounting for any ball that isn’t in the field of play.

Enter BP reader Brad McKay who e-mailed me and had a rather interesting way to look at the issue:

Following up on your work, I looked up the most frequent shifters since 2012 on FanGraphs. First, I ran a quick zero-order correlation with team HR/FB rate over the same period. Surprisingly (to me), there was a significant positive relationship between total batters faced with the shift and HR/FB rate. Just in case teams in hitter-friendly parks were more likely to shift, I controlled for each team's home HR park factor. Park factor was not significantly related to TBF with shift, and TBF with shift explained even more remaining variance in HR/FB after controlling for park factor.

Here's a summary for the results:

Zero-order correlation:

TBF with shift and HR/FB, r = .383, p = .03

Park factor, r = .721, p < .01.

Multiple Regression (partial correlation):

Park Factor, r = .728, p < .01

TBF with shift, r = .407, p = .03

Worth wondering if teams actually do give up more home runs when they shift... Maybe hitters are adding lift to their swing instead of hitting groundballs the other way.

He later wrote to say that he did not find any association between shifting and strikeouts or walks, using a similar methodology. Now, this most certainly isn’t bulletproof evidence that shifting increases the likelihood of a home run, but it is at least directional evidence. It could simply be a random fluke. Maybe there’s some other factor that links teams that shift with teams that give up a lot of home runs. Maybe the teams that aren’t very good tend to shift a lot. And give up a lot of home runs. But we do know that even the best-case version of The Shift leads to only very modest gains for the shifting team, and we have as yet not fully accounted for the balls that are not in play. If the home run finding is true, it could very easily wipe out the minor gains that come from The Shift’s magical BABIP properties.

The ‘Why’ Matters
If you look at The Shift in a vacuum, it would seem entirely obvious to work. If you assume that everything would work the same whether The Shift was on or not, you’d be a fool not to put The Shift on. But I’d submit that everything doesn’t work the same when The Shift is on. Paradoxically, understanding “why” might be the key to saving The Shift.

Let’s begin with the assumptions that pitchers are human and are aware of what’s going on in the infield behind them. While The Shift may be mathematically a great idea, it is also… well… weird. I think that if The Shift had been present since the 1920s, it would likely work just fine. But broadcasters still make a point of mentioning The Shift because it’s still a little strange. Notice that they never bother to say “two on the left, two on the right” because that’s the default that no one questions.

Now, put yourself inside the head of the pitcher. I doubt seriously that pitchers object to The Shift because they don’t get the basic underlying math. It’s not that hard to explain. But what’s harder to shake is the mental baseline that he has to deal with. He will always be able to look at a groundball and say “Well, against a normal infield, that would have been…” In some cases, the answer will be “that would have been an out.” (Boo.) In others, the answer would be “that would have been a hit“ (Hooray!).

People are loss-averse. The “pain” that comes from losing something that one “should” have had (the ball that squirts through the left side because Ortiz went oppo and poked it right past where the shortstop would have otherwise been) is greater than the “pleasure” that one gets from getting something that one did not previously have (the grounder that is hit right at the strangely placed shortstop that otherwise would have skipped into center field.) The fact is that both grounders represent roughly the same value (an out turned into a hit vs. a hit turned into an out), but the emotional value is higher in the first case.

And so, maybe without even noticing, pitchers might begin to shade their pitch selection to avoid anything that might result in an opposite-field hit. (This would make a fascinating study!) This sorta makes sense for practical reasons, although the entire reason that the Shift is on is that when the infield was played 2-and-2, the hitter didn’t normally go oppo.

Indeed, maybe the pitchers are just trying to be accommodating. The most obvious thing on the field when The Shift is on is… The Shift. It’s meant to help out when Ortiz hits a groundball, so I’d better go and get a groundball, rather than go get an out. I need to do this to help my manager justify himself for calling such a weird defensive formation. So much of pitching is in the element of surprise. Hitters are always looking for ways in which pitchers tip their pitches. In this case, they might fall into patterns that make them easier to read. Sometimes you don’t need to look for the slight tilt in the left arm. Just count the number of guys to the right of second base. What do you do when The Shift is actually a “tell?”

If teams want to continue shifting, then it will require a deep look under the hood to see whether their pitchers are falling prey to these natural inclinations. Are they prioritizing the wrong things (loss aversion? pitching for grounders rather than outs?) when coming up with a strategy to deal with this hitter? How to handle the very real (and yes, unjustifiably out-sized) sadness that comes from giving up a hit that a normal infield would have turned into an out. Not just on an after-the-fact basis, but for the fear of it happening. Essentially, you need pitchers who are willing to use all their tricks, even in a Shift situation and even the tricks that might be a little more likely to lead not just to a hit, but a hit that might be the fault of The Shift.

In 10 years, if shifting continues, and teams continue to introduce the idea of shifting even in their minor league affiliates, then perhaps we might see that the pitchers raised in this ecosystem won’t show some of these issues, but until then, it’s worth wondering whether teams shouldn’t shift their focus elsewhere in trying to scrape a strategic advantage. It’s not clear that The Shift is really doing much for teams now.

Russell A. Carleton is an author of Baseball Prospectus. Follow @pizzacutter4
Click here to see Russell's other articles. You can contact Russell by clicking here